Some problems when test own multiple pictures

[LiyingCV]

Greetings,
I refered the #28 and #60 , when I input one picture or multiple pictures the output voxel is a low quality. I check my code for long time but I still cannot resolve it. Follow is the main multiple pictures test code.

    IMG_SIZE = cfg.CONST.IMG_H, cfg.CONST.IMG_W
    CROP_SIZE = cfg.CONST.CROP_IMG_H, cfg.CONST.CROP_IMG_W

    test_transforms = utils.data_transforms.Compose([
        utils.data_transforms.CenterCrop(IMG_SIZE, CROP_SIZE),
        utils.data_transforms.RandomBackground(cfg.TEST.RANDOM_BG_COLOR_RANGE),
        utils.data_transforms.Normalize(mean=cfg.DATASET.MEAN, std=cfg.DATASET.STD),
        utils.data_transforms.ToTensor(),
    ])
    encoder = Encoder(cfg)
    decoder = Decoder(cfg)
    refiner = Refiner(cfg)
    merger = Merger(cfg)

    if torch.cuda.is_available():
        encoder = torch.nn.DataParallel(encoder).cuda()
        decoder = torch.nn.DataParallel(decoder).cuda()
        refiner = torch.nn.DataParallel(refiner).cuda()
        merger = torch.nn.DataParallel(merger).cuda()

    print('Loading weights from %s ...' % (cfg.CONST.WEIGHTS))
    checkpoint = torch.load(cfg.CONST.WEIGHTS)
    encoder.load_state_dict(checkpoint['encoder_state_dict'])
    decoder.load_state_dict(checkpoint['decoder_state_dict'])

    if cfg.NETWORK.USE_REFINER:
        refiner.load_state_dict(checkpoint['refiner_state_dict'])
        print('load Refiner')
    if cfg.NETWORK.USE_MERGER:
        merger.load_state_dict(checkpoint['merger_state_dict'])
        print('load Merger')

    encoder.eval()
    decoder.eval()
    refiner.eval()
    merger.eval()

    img1_path = './imgs/'
    for i in range(3):
        img1_np = []
        img1_np = cv2.imread(img1_path + '%d.png' % i)

    sample = np.array([img1_np])

    rendering_images = test_transforms(rendering_images=sample)

    rendering_images = rendering_images.unsqueeze(0)

    with torch.no_grad():
        image_features = encoder(rendering_images)
        raw_features, generated_volume = decoder(image_features)
        if cfg.NETWORK.USE_MERGER:
            generated_volume = merger(raw_features, generated_volume)
        else:
            generated_volume = torch.mean(generated_volume, dim=1)

        if cfg.NETWORK.USE_REFINER:
            generated_volume = refiner(generated_volume)

    generated_volume = generated_volume.squeeze(0)

    img_dir = './imgs/result'
    gv = generated_volume.data.cpu().numpy()
    rendering_views = utils.binvox_visualization.get_volume_views(gv, os.path.join(img_dir),1)
    voxel2obj('prediction.obj', gv > 0.4)

The input set:

Thank you in advance.

[hzxie]

Could you provide the results if you inference with each one of them?

[LiyingCV]

Could you provide the results if you inference with each one of them?

If I used one picture:

The result is:

Acturally, the result is stranges. I also save the .obj file, it result stranges too:

I try to do swap axes np.swapaxes(voxels, 2, 1) but it seems not work.

[hzxie]

It seems that there is something wrong with the single-view reconstruction result.
Please refer to issues #28 and #60 to solve it first.

[sujuyu]

Greetings,
I refered the #28 and #60 , when I input one picture or multiple pictures the output voxel is a low quality. I check my code for long time but I still cannot resolve it. Follow is the main multiple pictures test code.

    IMG_SIZE = cfg.CONST.IMG_H, cfg.CONST.IMG_W
    CROP_SIZE = cfg.CONST.CROP_IMG_H, cfg.CONST.CROP_IMG_W

    test_transforms = utils.data_transforms.Compose([
        utils.data_transforms.CenterCrop(IMG_SIZE, CROP_SIZE),
        utils.data_transforms.RandomBackground(cfg.TEST.RANDOM_BG_COLOR_RANGE),
        utils.data_transforms.Normalize(mean=cfg.DATASET.MEAN, std=cfg.DATASET.STD),
        utils.data_transforms.ToTensor(),
    ])
    encoder = Encoder(cfg)
    decoder = Decoder(cfg)
    refiner = Refiner(cfg)
    merger = Merger(cfg)

    if torch.cuda.is_available():
        encoder = torch.nn.DataParallel(encoder).cuda()
        decoder = torch.nn.DataParallel(decoder).cuda()
        refiner = torch.nn.DataParallel(refiner).cuda()
        merger = torch.nn.DataParallel(merger).cuda()

    print('Loading weights from %s ...' % (cfg.CONST.WEIGHTS))
    checkpoint = torch.load(cfg.CONST.WEIGHTS)
    encoder.load_state_dict(checkpoint['encoder_state_dict'])
    decoder.load_state_dict(checkpoint['decoder_state_dict'])

    if cfg.NETWORK.USE_REFINER:
        refiner.load_state_dict(checkpoint['refiner_state_dict'])
        print('load Refiner')
    if cfg.NETWORK.USE_MERGER:
        merger.load_state_dict(checkpoint['merger_state_dict'])
        print('load Merger')

    encoder.eval()
    decoder.eval()
    refiner.eval()
    merger.eval()

    img1_path = './imgs/'
    for i in range(3):
        img1_np = []
        img1_np = cv2.imread(img1_path + '%d.png' % i)

    sample = np.array([img1_np])

    rendering_images = test_transforms(rendering_images=sample)

    rendering_images = rendering_images.unsqueeze(0)

    with torch.no_grad():
        image_features = encoder(rendering_images)
        raw_features, generated_volume = decoder(image_features)
        if cfg.NETWORK.USE_MERGER:
            generated_volume = merger(raw_features, generated_volume)
        else:
            generated_volume = torch.mean(generated_volume, dim=1)

        if cfg.NETWORK.USE_REFINER:
            generated_volume = refiner(generated_volume)

    generated_volume = generated_volume.squeeze(0)

    img_dir = './imgs/result'
    gv = generated_volume.data.cpu().numpy()
    rendering_views = utils.binvox_visualization.get_volume_views(gv, os.path.join(img_dir),1)
    voxel2obj('prediction.obj', gv > 0.4)

The input set:

Thank you in advance.

Have you solved the problem?If so, can I refer to your solution